Generalized textured contact lens detection by extracting BSIF description from Cartesian iris images

Textured contact lenses cause severe problems for iris biometric systems because they can be used to alter the appearance of iris texture in order to deliberately increase the false positive and, especially, false negative match rates. Many texture analysis based techniques have been proposed for detecting the presence of cosmetic contact lenses. However, it has been shown recently that the generalization capability of the existing approaches is not sufficient because they have been developed for detecting specific lens texture patterns and evaluated only on those same lens types seen during development phase. This scenario does not apply in unpredictable practical applications because unseen lens patterns will be definitely experienced in operation. In this paper, we address this issue by studying the effect of different iris image preprocessing techniques and introducing a novel approach formore generalized cosmetic contact lens detection using binarized statistical image features (BSIF).Our extensive experimental analysis on benchmark datasets shows that the BSIF description extracted from preprocessed Cartesian iris texture images yields to promising generalization capabilities across unseen texture patterns and different iris sensors with mean equal error rate of 0.14%and 0.88%, respectively. The findings support the intuition that the textural differences between genuine iris texture and fake ones are best described by preserving the regular structure of different printing signatures without transforming the iris images into polar coordinate system.